> On Mon, Apr 08, 2002 at 02:15:06PM -0700, Jason R Thorpe wrote:
> 
>  > Ok.  An update on this.
> 
> ...and the plot thickens.
> 
>  > I submitted the necessary changes back to the FSF binutils folks, and
>  > got *significant* push-back.  They seem to strongly disagree with using
>  > the EI_OSABI field for this.
>  > 
>  > So, what should we do?  Should we back out these changes in our source
>  > tree?  If so, we should decide quick before the 1.6 branch is actually
>  > created.
> 
> I just noticed another problem with this.  The new elf32-arm-nbsd
> target doesn't grok our ELF core files, because they're dumped with
> EI_OSABI == 0 (the kernel doesn't really know any better).
> 
> Right now, I'm inclined to back out the EI_OSABI == ELFOSABI_NETBSD
> change, in light of the pushback from FSF and this gdb lossage.
> 

Agreed.  It means we loose the ability to tell an arm-linux object (.o 
file) from an arm-netbsd one, but that's life; and as was mentioned 
before, this is a temporary ABI.  Given that EI_OSABI is currently set by 
the GNU tools to an unauthorized value (ELFOSABI_ARM), this will certainly 
be fixed in the fully conforming ABI, so we can safely detect the 
differences.

R.